Non-Farm Entrepreneurship in Rural Africa: Patterns and ...ftp.iza.org/dp8008.pdf · Non-Farm...

DI

SC

US

SI

ON

P

AP

ER

S

ER

IE

S

Forschungsinstitut zur Zukunft der ArbeitInstitute for the Study of Labor

Non-Farm Entrepreneurship in Rural Africa:Patterns and Determinants

IZA DP No. 8008

February 2014

Paula NaglerWim Naudé

Non-Farm Entrepreneurship in Rural Africa:

Patterns and Determinants

Paula Nagler UNU-MERIT/Maastricht Graduate School of Governance

Wim Naudé

Maastricht School of Management, UNU-MERIT/MGSoG and IZA

Discussion Paper No. 8008 February 2014

IZA

P.O. Box 7240 53072 Bonn

Germany

Phone: +49-228-3894-0 Fax: +49-228-3894-180

E-mail: [email protected]

Any opinions expressed here are those of the author(s) and not those of IZA. Research published in this series may include views on policy, but the institute itself takes no institutional policy positions. The IZA research network is committed to the IZA Guiding Principles of Research Integrity. The Institute for the Study of Labor (IZA) in Bonn is a local and virtual international research center and a place of communication between science, politics and business. IZA is an independent nonprofit organization supported by Deutsche Post Foundation. The center is associated with the University of Bonn and offers a stimulating research environment through its international network, workshops and conferences, data service, project support, research visits and doctoral program. IZA engages in (i) original and internationally competitive research in all fields of labor economics, (ii) development of policy concepts, and (iii) dissemination of research results and concepts to the interested public. IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion. Citation of such a paper should account for its provisional character. A revised version may be available directly from the author.

mailto:[email protected]

IZA Discussion Paper No. 8008 February 2014

ABSTRACT

Non-Farm Entrepreneurship in Rural Africa: Patterns and Determinants

We are the first to provide a comparative empirical analysis of non-farm entrepreneurship in rural Africa, using the World Bank’s unique LSMSISA dataset. This dataset covers six countries over the period 2005 to 2012. We find that rural enterprises tend to be small, informal household enterprises that provide predominantly goods and services to the local economy, and operate intermittently due to seasonality in farming. We furthermore establish that the likelihood of operating an off-farm enterprise depends on individual capabilities, household characteristics and institutional factors. While the results of some variables show consistency across the sample, we also find much heterogeneity, suggesting that rural entrepreneurship is also a response to country-level circumstances and policies. Although more than 50 years have passed since rural development was identified as a priority for African countries, rural entrepreneurship continues to fulfill mainly a risk-diversifying role. This may suggest that policies to foster effective rural-urban migration and wage employment in rural areas, have largely failed in Africa.

NON-TECHNICAL SUMMARY This paper provides an empirical description of the patterns and determinants of non-farm entrepreneurship in rural Africa. It uses the World Bank’s LSMS-ISA dataset on six countries: Ethiopia, Niger, Nigeria, Malawi, Tanzania and Uganda. It finds that non-farm entrepreneur-ship predominantly creates informal jobs and mostly for family members; that these jobs tend to be transient and influenced by seasonality in agriculture; and that there is substantial heterogeneity across countries in the determinants of rural non-farm entrepreneurship. JEL Classification: Q12, O13, O55, M13, J43 Keywords: entrepreneurship, rural development, Sub-Saharan Africa, informal sector,

labour markets, small business, SMEs Corresponding author: Wim Naudé Maastricht School of Management PO Box 1203 6201 BE Maastricht The Netherlands E-mail: [email protected]

mailto:[email protected]

1 Introduction

Around a decade ago, Wiggens (2000) lamented the fact that “little isknown” about Africa’s rural non-farm economy, beyond an “embryonic set ofideas”. Studies based on consistent and cross-country comparable household-level data focusing on self-employment or entrepreneurship in the Africanrural non-farm economy have been limited, although we now know that thissector contributes with a relatively large and increasing share to householdincome in rural Africa (Davis et al., 2007; Reardon et al., 2006; Rijkers andCosta, 2012; Start, 2001). Since then the potential importance of the ru-ral non-farm sector has become firmly established, and the embryonic setof ideas mentioned by Wiggens elaborated in more detail by scholars. Thewhy, when and how of rural non-farm enterprises remain however relativelyunexplored from a comparative perspective. Most of the small (but grow-ing) scholarly literature makes use of one-period, single-country, and oftenrather limited survey data. Indeed, most empirical studies on entrepreneur-ship in Africa have focused on self-employment in urban areas. The neglectof a more solid empirical knowledge of rural entrepreneurship in Africa maybe a contributing factor in the limited success of rural development poli-cies, and of the fact that rural entrepreneurship does not feature in mostpoverty alleviation or entrepreneurship promotion strategies in Africa (Foxand Sohnesen, 2013).

The contribution of this paper lies in filling this gap by providing, on theone hand, comparative descriptive statistics on the patterns of rural non-farm enterprise characteristics; on the other hand, empirical estimates usingprobit regressions to identify individual, household and location variablesthat can explain the likelihood of rural households operating a non-farmenterprise. Our analysis is based on the recently available Living StandardsMeasurement Study - Integrated Surveys in Agriculture (LSMS-ISA) dataset that covers six countries over the period 2005 to 2012: Ethiopia, Malawi,Niger, Nigeria, Tanzania and Uganda.1

We find that rural non-farm enterprises are largely small and informal. Theyprovide predominantly goods and services to the local economy, and are oper-ated in most instances from within the household dwelling or the immediatesurroundings. They also reflect seasonality in farming, as only a share of en-

1 See also the World Bank’s website www.worldbank.org/lsms-isa

1

terprises are operated continuously throughout the year. The likelihood ofoperating an enterprise depends on individual capabilities, household char-acteristics and institutional factors. While the results of some variables showconsistency across the sample, we also find much heterogeneity, suggestingthat rural entrepreneurship is also a response to country-level circumstancesand policies. Although helping households to manage risks, overall ruralentrepreneurship does not seem to contribute significantly to employmentcreation, income growth and structural development in Africa.

The paper is organized as follows: in section 2 we discuss the relevant lit-erature and link it to our own empirical analysis. In section 3 we describethe database and the method used. In section 4 we present and discussour empirical findings on the patterns and determinants of rural non-farmentrepreneurship. The final section concludes.

2 Relevant Literature

2.1 Key Concepts

At the outset we clarify what we mean by rural non-farm entrepreneurship:self-employment in the rural non-farm economy.2 This is consistent withdefinitions and measures of entrepreneurship as the creation of a businessenterprise and working for one’s own account (Naude, 2011). With the ruralnon-farm economy we understand a residual sector wherein all non-farm ac-tivities in rural areas are captured, including agribusiness, services, trade andretail, tourism, rural industrialization, construction, and mining. Althoughdescribed as non-farm, many of these activities are linked to agriculture andcan actually take place on a farm (e.g. food processing, veterinary services)(Rijkers and Costa, 2012). Barrett et al. (2001) contain a more detaileddiscussion of the concepts of the rural economy in general, including agri-culture and the non-farm economy, and Roepstorff and Wiggens (2011) onthe concept and measurement of agribusiness.

2 There are many entrepreneurs in farming. In this paper however our focus is not onthe farmer-as-entrepreneur, but on the self-employment/entrepreneurship activities ofindividual household members in rural areas.

2

2.2 Importance of the Rural Non-Farm Economy

As we mentioned in the introduction, the why, when and how of rural non-farm enterprises in Africa remain relatively unexplored from a comparativeperspective. Until fairly recently a lack of comparable cross-country datalimited an empirical comparative analysis of the rural non-farm economy.The Rural Income Generating Activities (RIGA) database is a first effort toovercome this limitation. It has been constructed by the FAO and WorldBank from the LSMS for about 15 countries.3 As far as the patterns of therural non-farm economy are concerned, an analysis of the RIGA data baseby Davis et al. (2007) find that rural non-farm activities contributed morethan 50 percent to rural household income in 11 out of 15 countries, withan average across countries of 58 percent. They also find that the shareof non-farm income is however less important in Africa compared to otherregions.

This lack of empirical knowledge about the patterns and determinants ofrural entrepreneurship in Africa may be a contributing factor in the limitedsuccess of rural development policies and structural change in Africa, and ofthe fact that rural entrepreneurship does not feature in most poverty alle-viation or entrepreneurship promotion strategies (Fox and Sohnesen, 2013).Although data limitations undoubtedly are to blame, the neglect of ruralentrepreneurship in Africa can also be explained by the (earlier) thinkingon rural development in Africa, wherein the rural economy was expected tobecome less important as a contributor to rural household income over time(Lanjouw and Lanjouw, 2001). Urbanization and industrialization have beenconsequently a focus of development policies for much of the period sincethe 1960s, with the agricultural sector often (implicitly) repressed and reg-ulated (Havnevik et al., 2003). Most empirical studies on entrepreneurshipin Africa hence focus on urban areas.

That the rural non-farm entrepreneurship could potentially be importantfor development was only recognized from the late 1990s onwards, when itbecame clear that, contrary to expectations, the contribution of the sectorto rural household income did not decline over time, but in fact increased(Davis et al., 2007; Reardon et al., 2006). Today, the literature suggests that

3 The RIGA data set covers around 90,000 observations across Albania, Bangladesh, Bul-garia, Ecuador, Ghana, Guatemala, Indonesia, Madagascar, Malawi, Nepal, Nicaragua,Nigeria, Pakistan, Panama and Vietnam.

3

between 40 and 50 percent of household income in rural Africa originatesfrom rural non-farm entrepreneurship (Davis et al., 2007; Reardon et al.,2006; Rijkers and Costa, 2012; Start, 2001; Lanjouw and Lanjouw, 2001).For many of the countries under closer scrutiny in this paper, Reardon et al.(2006) report findings on the share of rural household income emanating fromthe non-farm economy. For instance in Ethiopia the share was 20 percent in1999, in Malawi 64 percent in 2004, and for Uganda and Tanzania 54 and46 percent in 2000, respectively.

Most of the empirical work on rural non-farm entrepreneurship confirmsthat it is a very heterogeneous sector (e.g. Davis and Bezemer, 2004; Barrettet al., 2001). Davis et al. (2007) find that most non-farm entrepreneurialactivities consist of very small and informal businesses in the commercial,trade and services sectors. The persistence of the smallness of firms in Africais noteworthy, as Haggblade et al. (1989) find that 95 percent of the ruralnon-farm enterprises employ less than five workers. In-depth studies on thepatterns and determinants of growth, and their start-up and failure rateshave however been hampered by a lack of panel data. The few studies thatexist have shed some light on the conditions that drive growth and failurein this sector over time.

Outside of Africa, Janvry and Sadoulet (2001) find that in Mexico the non-farm economy contributes on average with 55 percent to rural householdincome. Escobal (2001) reports a figure of 51 percent for Peru. Lanjouwand Lanjouw (2001) report figures of 39 percent for Brazil, 41 percent forChile, 50 percent for Colombia and 59 percent for Costa Rica. Shi et al.(2007) report 46 percent for China.

2.3 Determinants

In order to understand what determines the share of non-farm income inAfrica and why it is lower than elsewhere, we need to identify the deter-minants of (and hence obstacles to) rural non-farm entrepreneurship. Rel-atively few studies have dealt with this, and many of those that do, havebeen hampered by a lack of comparative and longitudinal data. Neverthe-less extant studies have made useful contributions, particularly in trying todisentangle the determinants of the decision of rural households to start anon-farm enterprise into push and pull factors, with household capabilities

4

and the institutional environment mediating (Reardon et al., 2006).

One of the most important push factors is the high degree of risk in Africanagriculture (Rijkers and Soderbom, 2013). Risk-averse farm households arekeen to diversify their income ex-ante in anticipation that a crop may fail; orex-post in the aftermath of a shock (Ackah, 2013). Rural entrepreneurshipas such is a form of self-insurance in the face of limited insurance markets.Longer-term structural push factors include surplus labor on farms, as grow-ing families put pressure on fixed parcels of farmland over time (Babatundeand Qaim, 2010; Reardon, 1997; Reardon et al., 2006). Push factors alsoinclude seasonality: household members may be pushed off-farm in the off-season. As such rural non-farm activities are secondary activities (Lanjouwand Lanjouw, 2001), and we expect to see a share of non-farm enterprisesthat do not operate continuously throughout the year.

Whether households start non-farm enterprises may also be due to their de-sire to utilize business opportunities, so called pull factors into entrepreneur-ship. In this regard household capabilities and assets, as well as individualcharacteristics have been found to be crucial (Barrett et al., 2001). House-hold capabilities and assets typically include gender, age (also a proxy forexperience), education, marital status (Abdulai and Delgado, 1999), as wellas financial assets (Ackah, 2013; Bhaumik et al., 2011), and the size of thehousehold itself.4 Women, at least in Africa, have been found to be morelikely to engage in the non-farm economy (Ackah, 2013; Canagarajah et al.,2001; Rijkers and Costa, 2012) or to migrate (Shi et al., 2007) compared tomen. Schooling seems to be relatively more important for finding non-farmwage employment than for starting a new business (Bayene, 2008; Daviset al., 2007; Elbers and Lanjouw, 2001; Reardon et al., 2006).

Both push and pull factors are influenced by the features of the local andregional economies where households are based (mostly exogenous for thehousehold). For instance the agro-climatic environment determines the ex-tent to which farming is productive and/or risky (Reardon, 1997). Thereis evidence that rural entrepreneurship fares better under favorable agro-climatic conditions (e.g. better rainfall) that are good for agricultural pro-ductivity and where other natural resources, e.g. in mines and tourist attrac-tions, can be found (Reardon et al., 2006). The location of the household also

4 The presence of children in a household may limit the choices of individual householdmembers, particularly women, within the non-farm economy, resulting e.g. in activitiesattached to the household residence (Havnevik et al., 2003).

5

determines the distance from urban areas, ports and markets. Fafchampsand Shilpi (2003) find that the share of non-farm wage employment declinesthe further a household lives away from a urban center, and that there is aU-shaped relationship between distance from a urban center and the shareof income from self-employment. This finding suggests some kind of protec-tion effect of deep rural isolation on non-farm enterprises. Also, the closera household is located to an urban center, the more likely an individualhousehold members migrates to this urban area (Reardon, 1997). Generallydistance, but also other determinants of market access such as the quality ofroads and utilities, can be important determinants of development in boththe farm and non-farm economy of Africa, and of the linkages between thetwo (Wiggens, 2000; Roepstorff and Wiggens, 2011).

2.4 Summary

Summarizing this section, we discussed the relevant literature on the patternsand determinants of rural non-farm entrepreneurship in Africa, and showedthat this literature has established that, contrary to earlier expectations, theshare of rural household income originating from the non-farm economy didnot decline over time, but in fact increased. As this share is around 40 percent in rural Africa, smaller than elsewhere in the world, it is important toidentify the determinants of rural household participation in the non-farmeconomy.

The existing literature has usefully made a distinction between push and pullfactors, and has found based mainly on individual country studies, push fac-tors such as shocks/risks, and surplus family labor to be significant, whereaspull factors such a individual and household level capabilities and assets,as well as institutional and regional features such as access to credit andinfrastructure, to be significant. In the remainder of the paper we explorethe LSMS-ISA dataset to investigate the patterns of non-farm entrepreneur-ship in rural Africa, and we use a probit regression analysis to identify thedeterminants of rural non-farm entrepreneurship. Based on the literature,we classify the determinants into shocks and risks, individual and householdlevel determinants, local and regional features and access to credit. Beforewe report these new empirical findings in section 4, we first explain thedataset and our estimation methods.

6

3 Data and Methodology

3.1 The Database

The LSMS-ISA database covers six countries in Sub-Saharan Africa: Ethi-opia, Malawi, Niger, Nigeria, Tanzania and Uganda.5 In all countries whereLSMS-ISA surveys have taken place, the plan is to conduct surveys ev-ery three years to obtain longitudinal data. At present a cross-sectionaldata base is available for all six countries, and panel data for Tanzania andUganda.

The questionnaire used has three sections: a community section, an agricul-tural section and a household section. The community section records accessto public services and infrastructure, social networks, governance, and retailprices. The agricultural section records crop production, storage and sales,land holdings, farming practices, input use and technology adaption, accessto and use of services, infrastructure and natural resources, livestock, andfishery. And the household questionnaire records household demographics,education, health and nutrition, food consumption and expenditure, non-food expenditure, employment, non-farm self-employment and other sourcesof income, dwelling conditions, durable assets, migration, and participationin projects and programs. The survey data has also been geo-referenced,which allows us to include the impact of geographically determined vari-ables such as rainfall, and distances from major roads and urban centersinto our regression analysis.

Although the data set is new, unique and facilitates the comparability be-tween countries (a feature that has so far been lacking in the available data),shortcomings remain that we need to acknowledge. For instance, while all sixcountries include a section with a set of questions covering the area of non-farm enterprises, the availability of certain questions differs between coun-tries and not all aspects are included in all country questionnaires. Questionsthat are available in all, or at least the majority of countries, often containdifferent answer possibilities of survey participants, also reflecting the spe-cific country context.

In the following section we use the LSMS-ISA data in two ways: first we

5 Data will also be collected for Mali in time.

7

explore the database to identify the salient patterns of rural non-farm en-trepreneurship (we describe this in section 4.1). Second, we run a numberof probit regressions to identify determinants of or factors associated withthe decision of households to participate in rural non-farm entrepreneurship(we report the results in section 4.2). Before proceeding, we use the nextsubsection to explain our estimation methods.

3.2 Estimation Method

We use probit regressions to identify the factors that determine or are as-sociated with the probability that a rural household operates a non-farmenterprise. Formally we estimate

Pr(Yi|Xi,Wi,Zi) = Φ(βXi + δWi + γZi)

where the dependent variable is a binary variable equal to one if a householdoperates a non-farm enterprise, and zero if not. Xi represents a vector ofindividual characteristics including a constant, and comprises the variablesgender, age, education (ability to read and write) and marital status ofthe household head. Wi represents a vector of household characteristicsincluding household size, the share of adults (household members age 15 orolder), the number of rooms in the dwelling as an approximation to householdwealth, and a binary variable if a household member has taken out credit overthe past 12 months, indicating the possibility of accessing financial support.Furthermore a binary variable if the household experienced food shortage ora shock over the past 12 months (idiosyncratic shocks, geographic shocks,institutional shocks and other shocks). Finally Zi represents a set of locationvariables, also named geo variables, including the household distance to thenext major road, the distance to the next population center, and annualprecipitation. These variables are based on our literature review.

8

4 Empirical Analysis

In section 4.1 we first explore the data and provide a set of tables and figureswith relevant patterns and characteristics of rural non-farm entrepreneur-ship, focusing on the extent and nature of entrepreneurial activities. Insection 4.2 we then present the regression results on the determinants ofparticipation.

4.1 Patterns of Rural Non-Farm Entrepreneurship

One of the established facts in the literature we highlighted in section 2 isthat non-farm entrepreneurship contributes a significant share to the incomeof rural households in Africa. The data from the LSMS-ISA is consistentwith this fact: non-farm entrepreneurship activities are widely prevalent inthe six African countries surveyed. As shown in Table 1 almost 42 percentout of the 24,551 rural households surveyed reported operating a non-farmenterprise (NFE in the table). Overall, our sample comprises 11,064 singlenon-farm enterprises in 8,137 rural households, resulting in an average of1.36 enterprises per household (evidence of portfolio entrepreneurship). Thecountry shares however vary widely, from as little as 17 percent in ruralMalawi, to almost 62 percent in rural Niger.

Table 1: Prevalence of rural non-farm enterprises

Country Nr of HH HH with in % in % Nr of Avg Nr ofsurveyed NFE weighted NFEs NFE/HH

Ethiopia 3,969 1,212 30.54 23.23 1,482 1.22Malawi 10,038 1,755 17.48 16.88 1,872 1.07Niger 2,430 1,427 58.72 61.73 2,188 1.53Nigeria 3,380 1,707 50.50 52.62 2,688 1.57Tanzania 2,629 1,083 41.19 39.10 1,363 1.26Uganda 2,105 953 45.27 42.24 1,471 1.54

Total 24,551 8,137 33.14 41.68 11,064 1.36

Source: Authors’ calculations based on LSMS-ISA data.Notes: weighted country shares are calculated using survey weights, total share in-cludes population weight.

As for the share of household income deriving from these enterprises, we only

9

have sufficient responses from Ethiopia. This indicates hat approximately27 percent of households engaged in non-farm entrepreneurship derive 50percent or more of their income from these activities, and approximately 5percent of households all income. This is consistent with previous estimates.

Using the Rural Income Generating Activities (RIGA) data set of the FAO,we derive shares of household income by household activity for four coun-tries: Malawi 2010/11, Niger 2011, Tanzania 2008/09, and Uganda 2010/11.Figure 1 shows the high importance of agricultural income and agriculturalwages in rural areas, and the share of self-employment: while it contributeswith less than 9 percent to total household income in Malawi, it representsas much as 36 percent in Niger, 16 in Tanzania and almost 21 percent inUganda. Comparing to urban areas (not included in Figure 1), the share ofagricultural income is decisively more important in rural areas, and incomederiving from self-employment smaller. In urban areas self-employment con-tributes with 22 percent in Malawi, 48 percent in Niger, 43 percent in Tan-zania and 33 percent in Uganda to household income (percentages rounded).

Figure 1: Contribution of activity to total household income (in %)

48.07

21.87

9.80

8.53

11.73

33.55

4.404.49

35.78

21.78

57.21

5.54

7.49

16.44

13.32

56.07

4.65

12.69

20.70

5.89

020

4060

8010

0P

erce

nt

Malawi Niger Tanzania Uganda

Agriculture Agricultural WageNon−Agricultural Wage Self−EmploymentTransfers & Other

Source: Authors’ calculations based on RIGA data (weighted shares).Note: Rural areas only.

The data furthermore casts doubt on the jobccreation potential of rural en-trepreneurship. The vast majority of non-farm enterprises surveyed are small

10

household enterprises: over 80 percent of the enterprise owners reported thatthey do not employ any non-household worker, and less than three percentemploy five or more non-household members. This suggests that non-farmenterprises are not significant drivers of wage employment in rural Africa. Itseems likely that non-farm enterprises fulfill a survival or risk-diversificationstrategy for rural households. This is also consistent with the existing lit-erature that we discussed in section 2. In addition we can mention thatDavis et al. (2007) find, using the RIGA data set, that most non-farm en-trepreneurial activities consist of very small and informal businesses in thecommercial, trade and services sectors, and Haggblade et al. (1989), usingcross-country data, that 95 percent of the rural non-farm enterprises employfive or less workers.

The assumption that rural non-farm entrepreneurship mainly fulfills a sur-vivalist function is further strengthened by the fact that between 91 andalmost 100 percent are operating informally. In Ethiopia, the country withthe highest rate of formality, only 8.95 percent of rural enterprises have alicense to operate, while in Niger only a very tiny 0.39 percent are registeredwith the government.6 Our findings confirm that non-farm entrepreneurshipis prominent in rural Africa, contributing to household activities and income,but with lower importance compared to urban areas. It furthermore doesnot contribute significantly to creation of wage employment in rural areas.

The data further shows that almost half of the non-farm business is operatedfrom inside the household’s residence or in the immediate surroundings (seeFigure 2). Rural non-farm enterprise activity is thus household-focused, asthe predominant employment of household members also suggests. Mostof the enterprise activity that is not in the surroundings of the householdresidence, is either located in a traditional market place or performed mobile.Less than 5 percent of rural non-farm enterprise activities are located ina commercial district or shop premises, emphasizing the largely informalnature of non-farm enterprise activities in rural Africa.

6 The definition of informal differs slightly between the various questionnaires. For in-stance in Ethiopia formal firms are those with a license to operate, in Malawi enterprisesregistered with the Malawi Revenue Authority, in Niger enterprises with a fiscal iden-tification number, in Nigeria whether the enterprise is registered with the government,and in Uganda whether the enterprise is registered for income tax and/or VAT. Therewere no questions in the Tanzania survey that recorded the formality of rural non-farmenterprises of households surveyed.

11

Figure 2: Place of business operation

47.38

20.54

12.80

6.474.77

1.75 0.99

5.10

010

2030

4050

Per

cent

Home Market

Mobile

Roadside

Commercial ShopClient

Industrial Site

Other

Source: Authors’ calculations based on LSMS-ISA data (weighted shares and populationweights).

Corresponding to these places of enterprise operation, most households in-dicate that their enterprises are providing consumer goods or services, orare engaged in trading. Between 20-30 percent of households indicate thattheir non-farm enterprises are engaged in processing of agricultural products.This is a surprising finding raising the question why agri-business is not moreprevalent. This question is reinforced when one considers the responses ofrural household enterprises as to the nature of their business clients. Theyreported that local final consumers is the most important customer base,accounting for 77 percent of sales. Most of the remainder of their sales areto markets, small businesses and traders in the local environment. Ruralnon-farm enterprises thus sell very little to governments, NGOs or manufac-turing firms, the main actors in most rural development support programs.Hence these programs seem overall largely irrelevant for most rural non-farmenterprises in Africa.

Although little research exists on the subject, it may be reasonable to sup-pose that rural entrepreneurship is affected by seasonality. For instance insection 2 we discussed that seasonal variations in agriculture can be a pushfactor for households to start and operate a non-farm enterprises during

12

the off-season. The impact of seasonality will therefore be reflected in thenumber of months per year a non-farm enterprise is in operation. Seasonalimpact may thus lead households to close their enterprises temporarily, forinstance when household members are required to work in agriculture. TheLSMS-ISA surveys capture the number of months per year a non-farm en-terprise was operating in the year preceding the survey. Figure 3 shows asignificant proportion of the rural enterprises operating less than six monthsper year. Only between 42 and 64 percent of all enterprises operate duringthe whole year, with the highest percentage found in Nigeria.

Figure 3: Months in operation

28.60

29.84

41.56

22.71

24.42

52.87

9.96

26.06

63.98

22.81

28.85

48.34

18.90

24.10

57.00

020

4060

8010

0P

erce

nt

Ethiopia Niger Nigeria Tanzania Uganda

<6 months 6−11 months12 months

Source: Authors’ calculations based on LSMS-ISA data (weighted shares).Notes: Malawi not reported due to lack of data. Enterprises that are less than one year

in operation are exluded.

From the survey we also calculated comparable shares for non-farm enter-prises operating in urban areas. As expected, we find that a higher percent-age of urban enterprises were in operation throughout the year. As Table 2indicates, there are proportionately less rural enterprises that operate for 12months during a year, as compared to urban enterprises. This suggests thatseasonality does have a potentially important influence on the dynamics ofrural entrepreneurship in Africa.

Summarizing this section we find that, although rural non-farm entrepreneur-

13

Table 2: Months of business operation - rural vs. urban (in %)

Niger Nigeria Tanzania Uganda

rural urban rural urban rural urban rural urban<6 23 9 10 7 23 12 19 76-11 24 16 26 24 29 25 24 1612 53 75 64 69 48 63 57 77

Source: Authors’ calculations based on LSMS-ISA data (weighted shares).

ship is prevalent and contributes with a non-negligible share to householdincome in Africa, it does not appear to be contributing much to the creationof wage employment, nor to structural change or agricultural value-added.It consists primarily of small, informal household enterprises operated byhousehold members. The business is generally operated from within or closeto the household residence, or sells products and offers services in traditionalmarkets, targeting local consumers. Seasonality has an impact, especiallywhen comparing with urban enterprises, as a decisive share of the businessesdoes not operate throughout the year, indicating that labor might be allo-cated to agricultural work during the planting and harvesting months. Gen-erally the descriptive statistics in this section is consistent with the previousliterature.

4.2 Determinants of Rural Non-Farm Entrepreneurship

In section 2 we discuss the determinants of rural entrepreneurship as iden-tified in the literature. In this section we set out the results of a probitregression analysis. In the first part we conduct the regressions using the ag-gregated sample and present also a cross-country analysis to identify sampleheterogeneity, in the second part we screen a set of country-specific variables.

4.2.1 Aggregated Analysis

First we run a set of probit regressions using the aggregated sample. Weinclude dummies for the different countries, with Ethiopia as the referencecategory, to have a first overview of variables that are significant in theAfrican context. The results are contained in Table 3. Second, we run

14

separate probit regressions for the six different countries of our sample toidentify the heterogeneity between countries. The results are contained inTable 4 for rural households, and in Appendix B for all households and urbanhouseholds.

Table 3 returns the coefficients and average marginal effects of the aggregateddata. The first and second column show the results of all households of thesample, the third and fourth column for rural households, and the fifthand sixth column for urban households. It shows that rural households inAfrica are on average almost 16 percentage points less likely compared tourban households to operate a non-farm enterprise. Cities are indeed morelikely places to find enterprises in Africa, as it is across the world. In thetable we also find significant country-level heterogeneity: the probability ofa household being involved in running a enterprise is almost 8 percentagepoints lower in Malawi than in Ethiopia, but 31 percentage points higherin Niger than in Ethiopia. According to Table 3 individual and householdlevel characteristics are important determinants of entrepreneurship overall,as reflected in the statistically significant coefficients on age, marital status,educational attainment (“read & write”) and household size. The signsand magnitudes of some of these coefficients are interesting. For instanceeducational attainment (“read & write”) is significant in both rural andurban areas, but with a positive effect in rural, and a negative effect in urbanareas. This reflects that educated individuals can find wage employmentmore easily in urban areas, but that this option for educated individuals islargely lacking in the rural non-farm economy in African countries.

As elsewhere in the literature push factors in the form of shocks and risks sig-nificantly affect rural entrepreneurship; the effects are however of a complexnature. Households that have experienced food shortages are 2.3 percent-age points more likely to operate a non-farm enterprise. “Other shocks” aresignificant for the aggregate sample and for rural households (and have a pos-itive impact suggesting a fallback option), geographical shocks are only sig-nificant for the aggregate sample (at 10 percent) and have a negative coeffi-cient, suggesting a lower likelihood of operating a non-farm enterprise. Also,distance to a road is a significant determinant for non-farm entrepreneurshipin rural areas with a negative impact (at 10 percent), but not significant forurban households. This result suggests that market access can constrainrural entrepreneurship and that investment in transport and transport in-frastructure may encourage rural entrepreneurship.

15

Table 3: Probit regressions - complete sample (coefficients and average marginaleffects)

Dependent (1) (2) (3) (4) (5) (6)NFE All Countries AME Rural AME Urban AME

Rural -0.430∗∗∗ -0.156∗∗∗

(0.06) (0.02)Female 0.121 0.040∗∗ 0.081 0.028 0.111 0.037

(0.08) (0.02) (0.05) (0.02) (0.10) (0.03)Rural x Female -0.009

(0.09)Age 0.014∗∗∗ -0.001∗∗∗ 0.010 -0.002∗∗∗ 0.023∗∗ -0.001

(0.01) (0.00) (0.01) (0.00) (0.01) (0.00)Age2 -0.000∗∗∗ -0.000∗∗∗ -0.000∗∗∗

(0.00) (0.00) (0.00)Married 0.100∗∗ 0.035∗∗ 0.035 0.012 0.164∗ 0.056∗

(0.05) (0.02) (0.05) (0.02) (0.09) (0.03)Read & Write 0.157∗∗∗ 0.055∗∗∗ 0.217∗∗∗ 0.077∗∗∗ -0.157∗∗ -0.052∗∗

(0.03) (0.01) (0.04) (0.01) (0.08) (0.02)HH Size 0.082∗∗∗ 0.021∗∗∗ 0.069∗∗∗ 0.020∗∗∗ 0.183∗∗∗ 0.032∗∗∗

(0.01) (0.00) (0.01) (0.00) (0.03) (0.01)HH Size2 -0.002∗∗∗ -0.001∗ -0.010∗∗∗

(0.00) (0.00) (0.00)Share of Adults -0.026 -0.009 0.109 0.038 -0.186 -0.063

(0.08) (0.03) (0.08) (0.03) (0.20) (0.07)Food Shortage 0.065∗ 0.023∗ 0.049 0.017 0.086 0.029

(0.04) (0.01) (0.04) (0.01) (0.08) (0.03)Shock (idiosyn.) -0.002 -0.001 0.009 0.003 -0.029 -0.010

(0.03) (0.01) (0.04) (0.01) (0.07) (0.02)Shock (geogr.) -0.067∗ -0.024∗ -0.055 -0.019 -0.131 -0.045

(0.04) (0.01) (0.04) (0.02) (0.10) (0.04)Shock (institut.) 0.005 0.002 0.005 0.002 0.003 0.001

(0.04) (0.01) (0.05) (0.02) (0.08) (0.03)Shock (other) 0.200∗∗ 0.070∗∗ 0.272∗∗∗ 0.097∗∗∗ 0.030 0.010

(0.09) (0.03) (0.10) (0.04) (0.20) (0.07)Dist to Road -0.006∗∗∗ -0.002∗∗ -0.005∗∗ -0.001∗ -0.008 -0.003

(0.00) (0.00) (0.00) (0.00) (0.01) (0.00)Dist to Road2 0.000∗∗∗ 0.000∗∗∗ -0.000

(0.00) (0.00) (0.00)Dist to Popcenter -0.001 -0.000 -0.002 -0.001 0.003 0.001

(0.00) (0.00) (0.00) (0.00) (0.01) (0.00)Dist to Popcenter2 0.000 0.000 -0.000

(0.00) (0.00) (0.00)Precipitation -0.000 -0.000 -0.000 -0.000 -0.000 -0.000

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)Malawi -0.220∗∗∗ -0.077∗∗∗ -0.233∗∗∗ -0.082∗∗∗ -0.654∗∗∗ -0.220∗∗∗

(0.05) (0.02) (0.06) (0.02) (0.12) (0.04)Niger 0.895∗∗∗ 0.312∗∗∗ 1.004∗∗∗ 0.352∗∗∗ -0.024 -0.008

(0.08) (0.03) (0.09) (0.03) (0.16) (0.05)Nigeria 0.824∗∗∗ 0.288∗∗∗ 0.810∗∗∗ 0.284∗∗∗ 0.411∗∗∗ 0.138∗∗∗

(0.06) (0.02) (0.07) (0.02) (0.13) (0.04)Tanzania 0.510∗∗∗ 0.178∗∗∗ 0.473∗∗∗ 0.166∗∗∗ 0.121 0.041

(0.06) (0.02) (0.07) (0.02) (0.12) (0.04)Uganda 0.465∗∗∗ 0.162∗∗∗ 0.451∗∗∗ 0.158∗∗∗ 0.069 0.023

(0.07) (0.02) (0.07) (0.02) (0.14) (0.05)cons -0.908∗∗∗ -1.225∗∗∗ -0.495

(0.18) (0.19) (0.38)N 31,664 31,664 23,882 23,882 7,782 7,782

Standard errors in parentheses. Survey weights included.∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01

Table 4 contains the results (average marginal effects) pertaining to ruralhouseholds broken down by country.7 It shows that in Nigeria women-headedhouseholds are almost 12 percentage points more likely to start a non-farmenterprise compared to a male-headed household, while in the other coun-tries gender of the household head is not a significant determinant. Age is asignificant determinant of rural entrepreneurship in most countries, and re-turns a small negative impact meaning that entrepreneurship in rural Africatends to be favored by younger households. The civil status (married) is as-sociated with an increase in the likelihood of non-farm entrepreneurship inNigeria, raising it by 11.8 percentage points, but with a decrease in the like-lihood in Tanzania by 4.6 percentage points. Being able to “read & write”is significant and positive in rural Malawi, Nigeria and Tanzania, suggest-ing that more educated household heads are more likely to start a non-farmenterprise in these countries. Household size is significant in four countries(Malawi, Nigeria, Tanzania and Uganda) with a positive coefficient, consis-tent with the findings elsewhere that larger households can allocate surpluslabor into non-farm entrepreneurship. The number of “rooms” (a proxy forhousehold wealth) is significant and positive for Ethiopia and Tanzania, andaccess to credit is significant and positive for Ethiopia and Malawi. Thesevariables are however endogenous and we have to read them as a positiveassociation.

Whether a household has been subject to food shortages over the past 12months (a proxy for agricultural productivity and perhaps household in-come) impacts the likelihood differently across countries. It is significantand negative for Malawi and Niger, limiting non-farm enterprises, but posi-tive and significant for Uganda. The results of shocks are, generally speak-ing, not significant in most countries, with some exceptions: idiosyncraticand geographic shocks return positive coefficients in Malawi, suggesting thathouseholds are pushed into non-farm enterprises in times of distress. Insti-tutional shocks are not significant in any country, and “other shocks” aresignificant and positive for Niger. Finally, Table 4 indicates that householddistance from a major road is significant for Nigeria and returns a lowerprobability of operating a non-farm enterprise (the probability decreases by0.2 percentage points per km of distance away from a road). Distance to thenext population center is significant for Malawi and Uganda, and equallydecreases the likelihood.

7 See Appendix B for the complete table including the regression coefficients.

17

Table 4: Probit regressions - rural households only (average marginal effects)

Dependent (1) (2) (3) (4) (5) (6)NFE Ethiopia Malawi Niger Nigeria Tanzania Uganda

Female -0.022 -0.004 -0.084 0.117∗∗ 0.046 0.002(0.03) (0.02) (0.07) (0.05) (0.03) (0.04)

Age -0.002∗∗∗ -0.001∗∗∗ 0.001 -0.001 -0.004∗∗∗ -0.004∗∗∗

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)Married -0.060 0.025 -0.047 0.118∗∗∗ -0.046∗ 0.056

(0.04) (0.02) (0.06) (0.05) (0.03) (0.04)Read & Write 0.036 0.041∗∗∗ 0.033 0.126∗∗∗ 0.073∗∗ 0.002

(0.02) (0.01) (0.03) (0.02) (0.03) (0.03)HH Size 0.001 0.010∗∗∗ -0.002 0.025∗∗∗ 0.018∗∗∗ 0.022∗∗∗

(0.01) (0.00) (0.01) (0.01) (0.01) (0.01)Share of Adults -0.125∗∗ -0.005 -0.140∗ 0.086 0.105∗ 0.205∗∗∗

(0.06) (0.03) (0.07) (0.05) (0.06) (0.07)Rooms 0.034∗∗∗ 0.006 0.005 0.029∗∗∗ 0.006

(0.01) (0.00) (0.01) (0.01) (0.01)Credit 0.089∗∗∗ 0.071∗∗∗ 0.033

(0.02) (0.01) (0.05)Food Shortage 0.031 -0.033∗∗∗ -0.094∗∗∗ 0.013 -0.005 0.070∗∗

(0.03) (0.01) (0.03) (0.02) (0.03) (0.04)Shock (idiosyn.) 0.036 0.062∗∗∗ 0.020 -0.040 0.022 0.030

(0.03) (0.01) (0.04) (0.03) (0.04) (0.03)Shock (geogr.) -0.040 0.021∗∗ -0.017 -0.052 0.008 0.030

(0.03) (0.01) (0.03) (0.03) (0.03) (0.03)Shock (institut.) 0.013 0.007 0.026 0.003 -0.020 -0.030

(0.03) (0.01) (0.03) (0.03) (0.03) (0.06)Shock (other) 0.122 0.018 0.090∗∗∗ 0.016 -0.064 0.083

(0.08) (0.04) (0.03) (0.09) (0.14) (0.09)Dist to Road 0.000 -0.000 -0.003 -0.002∗∗ -0.000 -0.004

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)Dist to Popcenter -0.000 -0.001∗∗∗ -0.001 -0.001 -0.001 -0.003∗

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)Precipitation 0.000 -0.000 0.000 -0.000 -0.000 -0.000

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)

N 3,320 10,017 2,430 3,324 2,956 1,789


18

We also screen the complete sample, broken down by countries, with adummy indicating if households are located in rural areas. We focus onthe dummy “rural” in the results of Table 8 (see Appendix B): the binaryvariable “rural” is highly significant for all countries with the exception ofNiger, and negative, indicating that the probability of households runninga non-farm enterprise is higher in urban areas of almost all countries, de-spite the fact that educated workers find easier wage employment in urbanareas. The difference in percentage points varies from approximately 10in Malawi and Uganda to almost 31 in Ethiopia. The coefficient of theinteraction “female x rural” is marginally significant in Ethiopia, where itdecreases the probability of operating a non-farm enterprise. Non-farm ruralentrepreneurs in Africa suffer in comparison to urban entrepreneurs in theirlack of amenities, densities and spillover effects, as well as in their greaterlack of wage employment alternatives. This was reflected in our findingsthat urban households are more likely to operate enterprises throughout theyear; and that being educated is more likely to lead to wage employment inurban areas.

Differences between the regressions pertaining to rural and urban house-holds (see Table 10) show the heterogeneity in determinants: while genderis significant in rural Nigeria and returns a positive result, only the averagemarginal effect in Ethiopia is marginally significant in the urban parts thecountry. In the urban areas age seems to be less important compared torural areas, where only the outcome for Niger is significant. Marital statusis significant in the rural areas of Nigeria and Tanzania with contradictingcoefficients, however in none of the urban areas. While the variable “read& write” increases the probability of engaging in non-farm entrepreneurshipin rural areas (where it is significant), the opposite is observed for urbanareas. As we have already mentioned, this possibly indicates better oppor-tunities of finding wage-employment in urban areas if the household head iseducated. The results for “household size” are similar in both tables: thecoefficient is positive and significant for most countries, suggesting the allo-cation of surplus labor into non-farm entrepreneurship. The share of adultsis significant in rural Ethiopia, Niger and Uganda, with contradicting effects,but only significant in urban Ethiopia with a positive coefficient. The num-ber of rooms has a positive and significant coefficient in rural Ethiopia andTanzania, while it is not significant in urban areas. Credit is positive and sig-nificant in rural Ethiopia and Malawi, and positive and significant in urbanEthiopia. Experiencing food shortages have a significant and negative effectin rural and urban Malawi and rural Niger, and a significant and positive

19

effect in rural Uganda. Experiencing a shock returns mixed results: in ruralMalawi an idiosyncratic and a geographic shock return positive coefficients,as well as other shocks in rural Niger. Idiosyncratic shocks return signifi-cant coefficients in urban Malawi and Uganda with contradicting signs. Ageographic shock is marginally significant for the average marginal effectsin urban Ethiopia, and institutional shocks return significant and negativeresults in urban Malawi and significant and positive results in urban Niger.The same applies to other shocks: in urban Tanzania the coefficient is sig-nificant and positive and in urban Uganda it is significant and negative.Distance to the nearest road is significant and negative in rural Nigeria,and urban Uganda. Distance to the nearest population center is significantand negative in rural Malawi and rural Uganda, but in none of the urbanareas. Precipitation is not significant in for all rural areas, but marginallysignificant and negative in urban Malawi and Nigeria.

4.2.2 Country-Specific Analysis

In this subsection we screen a number of country-specific variables that arenot available for all, or at least for the majority, of the countries, or a differ-ently coded per country. We focus on rural areas for the regressions (resultsare not reported, available upon request). We add a number of variables tothe base regressions in Table 4 country by country, and receive significantresults for the following countries and variables:

In Ethiopia the number of phones per household is significant at 5 percentand increases the probability of entrepreneurship by 3 percentage points peradditional phone, which also presents an indication for household wealth.The household’s religion is collectively significant at 1 percent. Taking thecountry’s traditional religion as the reference category, we find that being afollower of Islam increases the probability of a household starting a non-farmenterprise by almost 16 percentage points; and being a Christian householdincreases this probability by almost 10 percentage points. Other variables,as for example, the existence of a commercial bank or a micro-finance in-stitute in the community, the land use by households or the farm type, areinsignificant.

In Niger the variable “nomad”, indicating a nomadic lifestyle, decreases theprobability of operating a non-farm enterprise by 37 percentage points com-

20

pared to non-nomad households. While the possibility that vehicles passthrough the community throughout the year and the existence of electric-ity increases the likelihood of entrepreneurship (by 10 and 13 percentagepoints, respectively), the average marginal effect is negative for the numberof phones per household, the existence of a bank or micro-finance institute,or an irrigation system in the community. While the first two variablescould indicate a shift into other types of employment, an irrigation systemprobably results in more productive agricultural output, and labor could beallocated into agricultural work.

In Nigeria the number of beds per household is significant and increasesthe probability of entrepreneurship by 5 percentage point, per additionalbed. The variable “plot” is significant and indicates if the household ownsor cultivates a plot, and decreases the probability by almost 33 percentagepoints compared to households without a plot. The existence of a commercialbank or a micro-finance institute is however not significant in Nigeria.

In Tanzania the number of phones per household is significant and increasesthe probability of entrepreneurship by 7 percentage points per additionalphone. Other additional variables are not significant in rural Tanzania, forexample if households have livestock, or if they own or cultivate land.

In Uganda the number of phones per household is also significant and in-creases the probability of entrepreneurship by 4 percentage points per ad-ditional phone. Further significant variables, are the use of land, where theexistence of pasture or wetland increases the probability by 19 percentagepoints in both cases, compared to land cultivation by villagers. If house-holds cultivate land, the probability of entrepreneurship decreases by almost28 percentage points. If households own land is however not significant.

Summarizing this section we found positive impact of number of phonesand beds per households, with the exception of Niger, indicating a positiveassociation between household wealth and entrepreneurship. The existenceof a commercial bank and/or micro-finance institute in the community is notsignificant, or even results in negative marginal effects in the case of Niger.We also found that agriculture plays a role in some countries of our sample,for example in Niger via the irrigation system or in Uganda, where land useand land cultivation are significant. Other variables, for example, a nomadiclifestyle or religion, are country specific variables, and present habits andculture of a specific regional context.

21

5 Conclusion

The patterns and determinants of non-farm entrepreneurship in rural Africahave been neglected in the scholarly literature. In this paper we address thislacuna by providing new empirical insights using the World Bank’s LSMS-ISA data set which covers six Africa countries over the period 2005 - 2012.We find that rural non-farm enterprises are predominantly small, informalhousehold enterprises, operating from the immediate surroundings of thehousehold residence or in a traditional market, and provide mainly basicconsumer goods and services to the local economy. The majority of ruralnon-farm enterprises do not operate continuously over the year, reflectingthe impact of seasonality in agriculture on the allocation of household labor.

We furthermore establish that households operating a non-farm enterprisediffer in individual capabilities, household characteristics, and institutionalfactors, compared to household that are not engaged in entrepreneurship.We found that both push and pull factors matter for the household’s de-cision to operate a non-farm-enterprise. The effects of external shocks, theexperiencing of food shortages, the distance that households are located frommajor roads and cities, and the importance of gender and marital status areall difficult to generalize. In this regard we find significant heterogeneityacross the countries of our sample. This means that rural entrepreneurshipis responsive to country-level circumstances and policies.

What is the policy take-away from these empirical findings? It seems thatrural development policies in Sub-Saharan Africa have had little significantimpact in fostering the structural change in rural areas that are consistentwith notions of modern economic development. Being still a largely informaland survivalist sector, rural non-farm enterprises provide a risk-diversifyingmechanism for households, similar to the 1960s. More than 50 years laterrural non-farm entrepreneurship does still not contribute significantly to em-ployment creation, rural development or income growth. The increase in theshare of rural household income that emanates from this sector does notappear to be the outcome of successful policies, but the failure of policiesto foster effective rural-urban migration and wage employment. The factthat enterprises in rural Africa are still fulfilling a risk management functionby the year 2012, also suggests a sector that is reminiscent of economies atmuch lower gross national product (GNP) per capita. The comparativelyhigh economic growth rates in Africa over the past decade has not been

22

accompanied by significant structural economic change.

Although a risk-management function is important, given the shocks andseasonality that rural households in Africa have to contend with, we shouldnot expect a significant contribution from rural non-farm entrepreneurship toemployment creation and poverty reduction. This conclusion bears a resem-blance to conclusions by other scholars (e.g. Rijkers and Soderbom, 2013;Start, 2001) who have pointed out that Africa’s non-farm economy has alimited impact on poverty and growth. They have furthermore expressedconcern about its informal and segmented labor markets with low, exploita-tive wages, frequent discrimination against women, and regular employmentof children. Our results cannot, generally speaking, dispel such pessimisticevaluations of the state of non-farm entrepreneurship in rural Africa.

23

Acknowledgement

We are grateful to Luc Christiaensen for his inspiration and many discussionson the topic, to Bob Rijkers for useful comments on an earlier draft, toAmparo Palacios-Lopez for her assistance with the dataset and to JonathanKaminski and other participants in the World Bank’s “Telling Facts fromMyths - Agriculture in Africa” project for their comments and suggestions.The usual disclaimer applies.

24

References

Abdulai, A. and Delgado, C. (1999). Determinants of Non-Farm Earnings ofFarm-Based Husbands and Wives in Northern Ghana. American Journalof Agricultural Economics, 81 (1):117–130.

Ackah, C. (2013). Non-Farm Employment and Income in Rural Ghana.Journal of International Development, 25:325–339.

Babatunde, R. and Qaim, M. (2010). Impact of Off-Farm Income on FoodSecurity and Nutrition in Nigeria. Food Policy, 35:303–311.

Barrett, C., Reardon, T., and Webb, P. (2001). Non-Farm Income Diversi-fication and Household Livelihood Strategies in Rural Africa: Concepts,Issues, and Policy Implications. Food Policy, 26 (4):315–331.

Bayene, A. (2008). Determinants of Off-Farm Participation Decision of FarmHouseholds in Ethiopia. Agrekon, 47 (1):140–161.

Bhaumik, S., Dimova, R., and Nugent, J. (2011). Off-Farm Labor Supplyand Labor Markets in Rapidly Changing Circumstances: Bulgaria duringTransition. Economic Systems, 35:378–389.

Canagarajah, S., Newman, C., and Bhattamishra, R. (2001). Non-FarmIncome, Gender, and Inequality: Evidence from Rural Ghana and Uganda.Food Policy, 26 (4):405–420.

Davis, B., Winters, P., Carletto, G., Covarrubias, K., Quinones, E., Zezza,A., Stamoulis, K., Bonomi, G., and DiGuiseppe, S. (2007). Rural IncomeGenerating Activities: A Cross-Country Comparison. ESA Working PaperNo. 07-16, Rome: FAO.

Davis, J. and Bezemer, D. (2004). The Development of the Rural Non-FarmEconomy in Developing Countries and Transition Economies: Key Emerg-ing and Conceptual Issues. Chatham, UK: Natural Resources Institute.

Elbers, C. and Lanjouw, P. (2001). Intersectoral Transfer, Growth andInequality in Rural Ecuador. World Development, 29 (3):481–496.

Escobal, J. (2001). The Determinants of Non-Farm Income Diversificationin Rural Peru. World Development, 29 (3):497–508.

Fafchamps, M. and Shilpi, F. (2003). The Spatial Division of Labour inNepal. Journal of Development Studies, 39 (6):23–66.

25

Fox, L. and Sohnesen, T. (2013). Household Enterprises in Mozambique:Key to Poverty Reduction but not on the Development Agenda? PolicyResearch Working Paper 6570, Washington DC: The World Bank.

Haggblade, S., Hazell, P., and Brown, J. (1989). Farm-Non-Farm Linkagesin Rural Sub-Saharan Africa. World Development, 17 (8):1173–1201.

Havnevik, K., Hrsmar, M., and Sandstrm, E. (2003). Rural Developmentand the Private Sector in Sub-Saharan Africa: SIDA’s Experiences andApproaches in the 1990s. Stockholm: SIDA Evaluation 03/18.

Janvry, A. D. and Sadoulet, E. (2001). Income Strategies among RuralHouseholds in Mexico: The Role of Off-Farm Activities. World Develop-ment, 29 (3):467–480.

Lanjouw, J. O. and Lanjouw, P. (2001). The Rural Non-Farm Sector: Issuesand Evidence from Developing Countries. Agricultural Economics, 26:1–23.

Naude, W. (2011). Entrepreneurship is not a Binding Constraint on Growthand Development in the Poorest Countries. World Development, 39(1):33–44.

Reardon, T. (1997). Using Evidence of Household Income Diversificationto Inform Study of the Rural Non-Farm Labor Market in Africa. WorldDevelopment, 25 (5):735–747.

Reardon, T., Berdegue, J., Barrett, C., and Stamoulis, K. (2006). HouseholdIncome Diversification into Rural Non-Farm Activities. Baltimore: JohnsHopkins University Press.

Rijkers, B. and Costa, R. (2012). Gender and Rural Non-Farm Entrepreneur-ship. World Development, 40 (12):2411–2426.

Rijkers, B. and Soderbom, M. (2013). The Effects of Risk and Shocks on Non-Farm Enterprise Development in Rural Ethiopia. World Development,45:119–136.

Roepstorff, T. and Wiggens, S. (2011). New Global Realities GoverningAgribusiness. UNIDO: Vienna.

Shi, X., Heerink, N., and Qu, F. (2007). Choices between different Off-FarmEmployment Sub-Categories: An Empirical Analysis for Jiangxi Province,China. China Economic Review, 18:438–455.

26

Start, D. (2001). The Rise and Fall of the Rural Non-Farm Economy:Poverty Impacts and Policy Options. Development Policy Review, 19(4):491–505.

Wiggens, S. (2000). Interpreting Changes from the 1970s to the 1990sin African Agriculture through Village Studies. World Development, 28(4):631–662.

27

A Appendix

Summary statistics of section 4.2:

Table 5: Summary statistics probit model - overall

Household Ethiopia Malawi Niger Nigeria Tanzania Ugandaoperates a NFE No Yes No Yes No Yes No Yes No Yes No Yes

Rural 0.99 0.97 0.87 0.72 0.83 0.78 0.70 0.52 0.77 0.59 0.86 0.80

Female 0.21 0.19 0.25 0.18 0.13 0.10 0.18 0.14 0.27 0.26 0.32 0.27

Age 45.33 41.80 42.93 39.29 44.54 45.25 50.95 48.73 47.41 43.78 45.51 42.24

Married 0.81 0.82 0.72 0.81 0.88 0.90 0.73 0.82 0.53 0.52 0.66 0.77

Read & Write 0.40 0.49 0.62 0.76 0.32 0.34 0.58 0.71 0.69 0.81 0.66 0.68Household Size 5.04 5.35 4.49 5.00 6.05 6.80 4.87 5.69 4.85 5.40 5.86 6.71

Share of Adults 0.59 0.54 0.60 0.57 0.55 0.51 0.69 0.63 0.64 0.62 0.62 0.58

Rooms 1.64 1.82 2.50 2.75 2.55 2.76 3.22 3.44 2.76 2.81Credit 0.25 0.37 0.12 0.19 0.08 0.13

Food Shortage 0.32 0.37 0.49 0.41 0.41 0.33 0.28 0.32 0.20 0.20 0.21 0.24Shock (idiosyn.) 0.17 0.19 0.21 0.29 0.19 0.19 0.20 0.17 0.18 0.20 0.13 0.14Shock (geogr.) 0.24 0.22 0.44 0.40 0.38 0.33 0.11 0.07 0.17 0.15 0.26 0.30Shock (institut.) 0.34 0.37 0.43 0.43 0.30 0.35 0.10 0.09 0.22 0.25 0.05 0.06Shock (other) 0.02 0.04 0.02 0.02 0.14 0.19 0.01 0.02 0.01 0.01 0.02 0.02Dist to Road 14.98 14.93 9.70 7.29 13.60 8.89 14.47 10.84 17.15 14.52 7.78 7.00Dist to Popcenter 36.74 33.79 34.13 28.40 56.70 51.09 18.91 15.54 45.96 39.83 22.50 21.93Precipitation 1,166.20 1,206.96 1,067.49 1,043.17 387.09 406.95 1,514.74 1,475.12 1,060.38 1,058.18 1,244.05 1,232.79

N 2,757 1,212 9,699 2,572 1,543 2,425 2,185 2,815 2,145 1,857 1,409 1,307

Source: Authors’ calculations based on LSMS-ISA data (weighted shares). Bolded coefficents indicate differences between households with and without anon-farm enterprise that are significant at a 5 percent level.

28

Table 6: Summary statistics probit model - rural


Female 0.20 0.18 0.26 0.19 0.11 0.09 0.17 0.12 0.25 0.24 0.31 0.25

Age 45.37 41.77 43.50 39.84 44.19 44.96 51.38 48.98 48.95 44.99 46.28 42.33

Married 0.81 0.82 0.72 0.81 0.90 0.91 0.76 0.85 0.57 0.54 0.68 0.79


Share of Adults 0.59 0.54 0.59 0.55 0.53 0.50 0.65 0.60 0.61 0.58 0.60 0.57

Rooms 1.64 1.81 2.44 2.63 2.50 2.76 3.37 3.82 2.80 2.84Credit 0.26 0.37 0.11 0.19 0.08 0.11Food Shortage 0.32 0.37 0.52 0.47 0.47 0.38 0.26 0.27 0.21 0.21 0.22 0.27Shock (idiosyn.) 0.17 0.19 0.22 0.31 0.19 0.19 0.21 0.18 0.15 0.16 0.13 0.16Shock (geogr.) 0.24 0.22 0.48 0.52 0.45 0.40 0.14 0.11 0.18 0.17 0.29 0.35

Shock (institut.) 0.34 0.38 0.45 0.50 0.32 0.36 0.11 0.10 0.21 0.21 0.05 0.05Shock (other) 0.02 0.04 0.02 0.02 0.15 0.21 0.01 0.01 0.01 0.00 0.02 0.02Dist to Road 14.98 14.97 10.84 9.37 16.26 11.09 18.33 16.75 20.80 21.16 8.55 8.11Dist to Popcenter 36.69 33.44 37.69 35.49 65.86 61.96 23.89 22.92 53.91 52.77 24.78 25.31Precipitation 1,166.26 1,206.95 1,068.41 1,049.48 376.69 401.14 1,470.53 1,417.18 1,071.51 1,065.28 1,238.64 1,224.92

N 2,547 919 8,283 1,755 1,003 1,427 1,673 1,707 1,584 1,083 1,152 953

Source: Authors’ calculations based on LSMS-ISA data (weighted shares). Bolded coefficents indicate differences between households with and without anon-farm enterprise that are significant at a 5 percent level.

29

Table 7: Summary statistics probit model - urban


Female 0.30 0.38 0.19 0.15 0.18 0.15 0.21 0.17 0.34 0.29 0.35 0.37Age 37.43 42.87 39.00 37.86 46.19 46.30 49.97 48.46 41.70 42.25 40.63 41.91Married 0.62 0.67 0.71 0.80 0.76 0.84 0.66 0.79 0.39 0.51 0.52 0.69


Share of Adults 0.73 0.64 0.66 0.61 0.63 0.57 0.77 0.66 0.73 0.68 0.75 0.64

Rooms 2.07 2.30 2.87 3.08 2.80 2.78 2.65 2.92 2.44 2.69Credit 0.13 0.26 0.19 0.20 0.12 0.18Food Shortage 0.23 0.20 0.29 0.25 0.13 0.15 0.33 0.38 0.17 0.20 0.17 0.13Shock (idiosyn.) 0.14 0.15 0.15 0.25 0.19 0.20 0.16 0.16 0.26 0.26 0.15 0.09Shock (geogr.) 0.09 0.05 0.11 0.10 0.07 0.06 0.03 0.03 0.13 0.12 0.08 0.11Shock (institut.) 0.18 0.20 0.28 0.25 0.25 0.33 0.09 0.08 0.25 0.29 0.03 0.06Shock (other) 0.02 0.01 0.02 0.02 0.10 0.11 0.02 0.03 0.00 0.02 0.02 0.01Dist to Road 15.21 13.53 1.81 1.88 1.05 0.92 5.57 4.50 4.86 4.75 2.45 2.84Dist to Popcenter 45.04 46.89 9.46 9.96 13.39 11.71 7.43 7.64 18.94 21.18 6.68 9.29Precipitation 1,155.65 1,207.70 1,061.12 1,026.76 436.25 428.00 1,616.71 1,537.23 1,028.66 1,049.29 1,281.45 1,262.21

N 210 293 1,416 817 540 998 512 1,108 655 876 257 354

Source: Authors’ calculations based on LSMS-ISA data (weighted shares). Bolded coefficents indicate differences between households with and withouta non-farm enterprise that are significant at a 5 percent level.

30

B Appendix

Table 8: Probit regressions - all households (coefficients and average marginal effects)

Dependent (1) (2) (3) (4) (5) (6)NFE Ethiopia AME Malawi AME Niger AME

Rural -0.808∗∗∗ -0.309∗∗∗ -0.357∗∗∗ -0.103∗∗∗ 0.139 0.050(0.14) (0.05) (0.07) (0.02) (0.15) (0.05)

Female 0.264 -0.019 0.010 0.002 -0.117 -0.048(0.17) (0.03) (0.15) (0.02) (0.14) (0.05)

Rural x Female -0.337∗ -0.004 -0.017(0.18) (0.15) (0.15)

Age -0.011 -0.002∗∗∗ 0.004 -0.001∗∗∗ 0.022∗ 0.001(0.01) (0.00) (0.01) (0.00) (0.01) (0.00)

Age2 0.000 -0.000∗ -0.000∗

(0.00) (0.00) (0.00)Married -0.195 -0.059 0.122∗∗ 0.031∗∗ 0.016 0.006

(0.13) (0.04) (0.06) (0.02) (0.13) (0.05)Read & Write 0.121 0.035 0.153∗∗∗ 0.039∗∗∗ -0.036 -0.013

(0.08) (0.02) (0.04) (0.01) (0.07) (0.03)HH Size -0.068 0.002 0.078∗∗ 0.013∗∗∗ -0.026 0.004

(0.07) (0.01) (0.03) (0.00) (0.03) (0.01)HH Size2 0.007 -0.003 0.003∗

(0.01) (0.00) (0.00)Share of Adults -0.427∗∗ -0.123∗∗ 0.025 0.007 -0.370∗∗ -0.134∗∗

(0.20) (0.06) (0.11) (0.03) (0.16) (0.06)Rooms 0.115∗∗∗ 0.033∗∗∗ 0.031∗ 0.008∗∗ -0.009 -0.003

(0.04) (0.01) (0.02) (0.00) (0.03) (0.01)Credit 0.297∗∗∗ 0.090∗∗∗ 0.202∗∗∗ 0.056∗∗∗

(0.07) (0.02) (0.05) (0.02)Food Shortage 0.104 0.030 -0.144∗∗∗ -0.037∗∗∗ -0.223∗∗∗ -0.081∗∗∗

(0.09) (0.03) (0.03) (0.01) (0.06) (0.02)Shock (idiosyn.) 0.116 0.035 0.275∗∗∗ 0.076∗∗∗ 0.057 0.021

(0.08) (0.02) (0.04) (0.01) (0.09) (0.03)Shock (geogr.) -0.145 -0.041 0.088∗∗ 0.023∗∗ -0.079 -0.029

(0.12) (0.03) (0.04) (0.01) (0.08) (0.03)Shock (institut.) 0.047 0.014 -0.012 -0.003 0.117 0.042

(0.10) (0.03) (0.04) (0.01) (0.07) (0.03)Shock (other) 0.372∗ 0.120 0.031 0.008 0.221∗∗∗ 0.078∗∗∗

(0.21) (0.07) (0.13) (0.04) (0.08) (0.03)Dist to Road 0.000 0.000 -0.001 -0.000 -0.009 -0.004

(0.01) (0.00) (0.01) (0.00) (0.01) (0.00)Dist to Road2 0.000 -0.000 -0.000

(0.00) (0.00) (0.00)Dist to Popcenter -0.001 -0.000 -0.010∗∗∗ -0.001∗∗∗ -0.004 -0.001

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)Dist to Popcenter2 -0.000 0.000∗∗∗ 0.000

(0.00) (0.00) (0.00)Precipitation 0.000 0.000 -0.000∗∗ -0.000∗∗ 0.000 0.000

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)cons 0.544 -0.784∗∗∗ 0.048

(0.47) (0.22) (0.42)N 3,797 3,797 12,246 12,246 3,968 3,968


31

(7) (8) (9) (10) (11) (12)Nigeria AME Tanzania AME Uganda AME

Rural -0.386∗∗∗ -0.140∗∗∗ -0.577∗∗∗ -0.204∗∗∗ -0.235∗∗ -0.097∗∗

(0.09) (0.03) (0.09) (0.03) (0.12) (0.04)Female 0.259∗ 0.090∗∗∗ 0.001 0.038 0.082 0.001

(0.14) (0.03) (0.10) (0.03) (0.17) (0.03)Rural x Female -0.001 0.149 -0.092

(0.15) (0.12) (0.18)Age 0.023∗∗ -0.001 0.010 -0.003∗∗∗ 0.035∗∗ -0.003∗∗∗

(0.01) (0.00) (0.01) (0.00) (0.01) (0.00)Age2 -0.000∗∗∗ -0.000∗∗ -0.000∗∗∗

(0.00) (0.00) (0.00)Married 0.251∗∗∗ 0.092∗∗∗ -0.041 -0.015 0.134 0.049

(0.09) (0.03) (0.06) (0.02) (0.11) (0.04)Read & Write 0.181∗∗∗ 0.066∗∗∗ 0.208∗∗∗ 0.077∗∗∗ 0.010 0.004

(0.06) (0.02) (0.08) (0.03) (0.07) (0.03)HH Size 0.100∗∗∗ 0.024∗∗∗ 0.077∗∗∗ 0.023∗∗∗ 0.135∗∗∗ 0.023∗∗∗

(0.03) (0.00) (0.02) (0.01) (0.03) (0.00)HH Size2 -0.003∗∗ -0.001∗∗ -0.005∗∗∗

(0.00) (0.00) (0.00)Share of Adults -0.042 -0.015 0.136 0.050 0.377∗∗ 0.137∗∗

(0.14) (0.05) (0.14) (0.05) (0.17) (0.06)Rooms 0.057∗∗∗ 0.021∗∗∗ 0.014 0.005

(0.02) (0.01) (0.02) (0.01)Credit 0.153∗ 0.057∗

(0.09) (0.03)Food Shortage 0.068 0.024 0.034 0.012 0.149∗ 0.055∗

(0.06) (0.02) (0.06) (0.02) (0.09) (0.03)Shock (idiosyn.) -0.071 -0.026 0.019 0.007 -0.014 -0.005

(0.06) (0.02) (0.07) (0.03) (0.08) (0.03)Shock (geogr.) -0.160∗ -0.058∗ -0.020 -0.008 0.103 0.038

(0.08) (0.03) (0.07) (0.03) (0.08) (0.03)Shock (institut.) -0.036 -0.013 0.001 0.000 -0.034 -0.012

(0.07) (0.03) (0.07) (0.03) (0.14) (0.05)Shock (other) 0.012 0.004 0.275 0.102 0.035 0.013

(0.19) (0.07) (0.28) (0.10) (0.22) (0.08)Dist to Road -0.009∗ -0.003∗∗ -0.004 -0.001 -0.044∗∗∗ -0.007∗∗

(0.00) (0.00) (0.00) (0.00) (0.02) (0.00)Dist to Road2 0.000 0.000 0.002∗∗∗

(0.00) (0.00) (0.00)Dist to Popcenter -0.000 -0.000 0.001 -0.000 -0.018∗∗ -0.002∗

(0.01) (0.00) (0.00) (0.00) (0.01) (0.00)Dist to Popcenter2 -0.000 -0.000 0.000∗∗

(0.00) (0.00) (0.00)Precipitation -0.000∗ -0.000∗ 0.000 0.000 -0.000 -0.000

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)cons -0.452 -0.526∗ -0.944∗∗

(0.28) (0.29) (0.46)N 4,914 4,914 4,419 4,419 2,268 2,268

32

Table 9: Probit regressions - rural households only (coefficients and averagemarginal effects)

(1) (2) (3) (4) (5) (6)Ethiopia AME Malawi AME Niger AME

Female -0.078 -0.022 -0.017 -0.004 -0.226 -0.084(0.12) (0.03) (0.07) (0.02) (0.18) (0.07)

Age -0.012 -0.002∗∗∗ 0.002 -0.001∗∗∗ 0.032∗∗ 0.001(0.01) (0.00) (0.01) (0.00) (0.01) (0.00)

Age2 0.000 -0.000 -0.000∗∗

(0.00) (0.00) (0.00)Married -0.200 -0.060 0.107 0.025 -0.131 -0.047

(0.13) (0.04) (0.07) (0.02) (0.18) (0.06)Read & Write 0.125 0.036 0.174∗∗∗ 0.041∗∗∗ 0.093 0.033

(0.08) (0.02) (0.04) (0.01) (0.09) (0.03)HH Size -0.076 0.001 0.061∗ 0.010∗∗∗ -0.059 -0.002

(0.07) (0.01) (0.03) (0.00) (0.05) (0.01)HH Size2 0.008 -0.002 0.004

(0.01) (0.00) (0.00)Share of Adults -0.433∗∗ -0.125∗∗ -0.021 -0.005 -0.388∗ -0.140∗

(0.21) (0.06) (0.11) (0.03) (0.20) (0.07)Rooms 0.117∗∗∗ 0.034∗∗∗ 0.026 0.006 0.013 0.005

(0.04) (0.01) (0.02) (0.00) (0.04) (0.01)Credit 0.295∗∗∗ 0.089∗∗∗ 0.268∗∗∗ 0.071∗∗∗

(0.07) (0.02) (0.05) (0.01)Food Shortage 0.107 0.031 -0.136∗∗∗ -0.033∗∗∗ -0.256∗∗∗ -0.094∗∗∗

(0.09) (0.03) (0.04) (0.01) (0.07) (0.03)Shock (idiosyn.) 0.120 0.036 0.241∗∗∗ 0.062∗∗∗ 0.054 0.020

(0.08) (0.03) (0.05) (0.01) (0.10) (0.04)Shock (geogr.) -0.143 -0.040 0.086∗∗ 0.021∗∗ -0.048 -0.017

(0.12) (0.03) (0.04) (0.01) (0.08) (0.03)Shock (institut.) 0.045 0.013 0.030 0.007 0.072 0.026

(0.10) (0.03) (0.04) (0.01) (0.09) (0.03)Shock (other) 0.378∗ 0.122 0.072 0.018 0.256∗∗∗ 0.090∗∗∗

(0.21) (0.08) (0.15) (0.04) (0.09) (0.03)Dist to Road 0.001 0.000 -0.001 -0.000 -0.008 -0.003

(0.01) (0.00) (0.01) (0.00) (0.01) (0.00)Dist to Road2 0.000 -0.000 0.000

(0.00) (0.00) (0.00)Dist to Popcenter -0.001 -0.000 -0.011∗∗∗ -0.001∗∗∗ -0.005 -0.001

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)Dist to Popcenter2 -0.000 0.000∗∗∗ 0.000

(0.00) (0.00) (0.00)Precipitation 0.000 0.000 -0.000 -0.000 0.001 0.000

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)cons -0.219 -1.053∗∗∗ 0.004

(0.45) (0.23) (0.56)N 3,320 3,320 10,017 10,017 2,430 2,430


33


Female 0.319∗∗ 0.117∗∗ 0.124 0.046 0.005 0.002(0.13) (0.05) (0.09) (0.03) (0.10) (0.04)

Age 0.018 -0.001 0.000 -0.004∗∗∗ 0.037∗∗ -0.004∗∗∗

(0.01) (0.00) (0.01) (0.00) (0.02) (0.00)Age2 -0.000∗∗ -0.000 -0.001∗∗∗

(0.00) (0.00) (0.00)Married 0.309∗∗ 0.118∗∗∗ -0.124∗ -0.046∗ 0.155 0.056

(0.12) (0.05) (0.07) (0.03) (0.12) (0.04)Read & Write 0.329∗∗∗ 0.126∗∗∗ 0.201∗∗ 0.073∗∗ 0.007 0.002

(0.07) (0.02) (0.08) (0.03) (0.08) (0.03)HH Size 0.083∗∗∗ 0.025∗∗∗ 0.059∗∗∗ 0.018∗∗∗ 0.118∗∗∗ 0.022∗∗∗

(0.03) (0.01) (0.02) (0.01) (0.03) (0.01)HH Size2 -0.001 -0.001 -0.004∗∗∗

(0.00) (0.00) (0.00)Share of Adults 0.228 0.086 0.287 0.105∗ 0.567∗∗∗ 0.205∗∗∗

(0.14) (0.05) (0.17) (0.06) (0.19) (0.07)Rooms 0.079∗∗∗ 0.029∗∗∗ 0.017 0.006

(0.02) (0.01) (0.03) (0.01)Credit 0.089 0.033

(0.12) (0.05)Food Shortage 0.034 0.013 -0.012 -0.005 0.192∗ 0.070∗∗

(0.06) (0.02) (0.08) (0.03) (0.10) (0.04)Shock (idiosyn.) -0.106 -0.040 0.059 0.022 0.083 0.030

(0.07) (0.03) (0.10) (0.04) (0.09) (0.03)Shock (geogr.) -0.138 -0.052 0.021 0.008 0.082 0.030

(0.09) (0.03) (0.09) (0.03) (0.09) (0.03)Shock (institut.) 0.009 0.003 -0.055 -0.020 -0.083 -0.030

(0.09) (0.03) (0.09) (0.03) (0.16) (0.06)Shock (other) 0.041 0.016 -0.181 -0.064 0.226 0.083

(0.25) (0.09) (0.41) (0.14) (0.23) (0.09)Dist to Road -0.009∗ -0.002∗∗ -0.003 -0.000 -0.038∗∗ -0.004

(0.00) (0.00) (0.00) (0.00) (0.02) (0.00)Dist to Road2 0.000 0.000 0.002∗∗∗

(0.00) (0.00) (0.00)Dist to Popcenter -0.004 -0.001 -0.002 -0.001 -0.022∗∗ -0.003∗

(0.01) (0.00) (0.00) (0.00) (0.01) (0.00)Dist to Popcenter2 0.000 0.000 0.000∗∗

(0.00) (0.00) (0.00)Precipitation -0.000 -0.000 -0.000 -0.000 -0.000 -0.000

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)cons -0.990∗∗∗ -0.676∗∗ -1.231∗∗

(0.34) (0.34) (0.52)N 3,324 3,324 2,956 2,956 1,789 1,789

34

Table 10: Probit regressions - urban households only (coefficients and averagemarginal effects)

(1) (2) (3) (4) (5) (6)Ethiopia AME Malawi AME Niger AME

Female 0.433 0.147∗ 0.056 0.020 0.003 0.001(0.26) (0.08) (0.16) (0.06) (0.17) (0.06)

Age -0.004 0.001 0.021 -0.002 -0.015 -0.002∗

(0.02) (0.00) (0.01) (0.00) (0.02) (0.00)Age2 0.000 -0.000∗∗ 0.000

(0.00) (0.00) (0.00)Married -0.057 -0.020 0.127 0.044 0.230 0.079

(0.20) (0.07) (0.13) (0.04) (0.16) (0.06)Read & Write -0.175 -0.061 -0.070 -0.025 -0.459∗∗∗ -0.155∗∗∗

(0.22) (0.08) (0.15) (0.05) (0.10) (0.03)HH Size 0.304∗∗ 0.078∗∗∗ 0.137∗ 0.035∗∗∗ 0.058 0.020∗∗∗

(0.14) (0.02) (0.07) (0.01) (0.04) (0.01)HH Size2 -0.011 -0.004 0.000

(0.01) (0.01) (0.00)Share of Adults 0.593∗ 0.205∗ 0.189 0.066 -0.239 -0.080

(0.34) (0.12) (0.24) (0.08) (0.25) (0.08)Rooms 0.012 0.004 0.028 0.010 -0.029 -0.010

(0.07) (0.02) (0.03) (0.01) (0.03) (0.01)Credit 0.670∗∗∗ 0.221∗∗∗ -0.003 -0.001

(0.21) (0.06) (0.14) (0.05)Food Shortage -0.122 -0.043 -0.152∗ -0.053∗ -0.030 -0.010

(0.24) (0.09) (0.09) (0.03) (0.14) (0.05)Shock (idiosyn.) -0.034 -0.012 0.425∗∗∗ 0.156∗∗∗ 0.031 0.010

(0.24) (0.08) (0.08) (0.03) (0.11) (0.04)Shock (geogr.) -0.542 -0.190∗ -0.000 -0.000 -0.166 -0.057

(0.33) (0.11) (0.13) (0.05) (0.16) (0.06)Shock (institut.) 0.157 0.054 -0.211∗∗ -0.073∗∗ 0.182∗∗ 0.060∗∗

(0.12) (0.04) (0.11) (0.04) (0.09) (0.03)Shock (other) -0.743 -0.255 -0.181 -0.061 -0.029 -0.010

(0.62) (0.20) (0.24) (0.08) (0.13) (0.04)Dist to Road 0.005 0.000 -0.006 -0.001 0.011 -0.012

(0.01) (0.00) (0.04) (0.01) (0.10) (0.02)Dist to Road2 -0.000 0.000 -0.024∗

(0.00) (0.00) (0.01)Dist to Popcenter 0.005 0.001 -0.006 -0.001 -0.001 -0.000

(0.01) (0.00) (0.01) (0.00) (0.01) (0.00)Dist to Popcenter2 -0.000 0.000 -0.000

(0.00) (0.00) (0.00)Precipitation 0.000 0.000 -0.001∗ -0.000∗ -0.001 -0.000

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)cons -1.695∗∗ -0.796 1.122∗∗

(0.70) (0.58) (0.50)N 477 477 2,229 2,229 1,535 1,535


35


Female 0.113 0.035 0.100 0.036 0.094 0.032(0.16) (0.05) (0.11) (0.04) (0.19) (0.06)

Age 0.027∗ -0.001 0.009 -0.001 0.033 0.002(0.02) (0.00) (0.02) (0.00) (0.03) (0.00)

Age2 -0.000∗∗ -0.000 -0.000(0.00) (0.00) (0.00)

Married 0.182 0.060 0.080 0.029 0.122 0.042(0.14) (0.05) (0.12) (0.04) (0.20) (0.07)

Read & Write -0.236∗∗ -0.073∗∗ 0.255 0.095 0.004 0.001(0.10) (0.03) (0.18) (0.07) (0.20) (0.07)

HH Size 0.204∗∗∗ 0.030∗∗∗ 0.232∗∗∗ 0.050∗∗∗ 0.187∗∗∗ 0.029∗∗∗

(0.05) (0.01) (0.06) (0.01) (0.05) (0.01)HH Size2 -0.012∗∗∗ -0.011∗∗∗ -0.009∗∗∗

(0.00) (0.00) (0.00)Share of Adults -0.280 -0.090 0.235 0.085 -0.522 -0.177

(0.27) (0.09) (0.27) (0.10) (0.40) (0.14)Rooms -0.001 -0.001 -0.022 -0.007

(0.03) (0.01) (0.05) (0.02)Credit 0.201 0.071

(0.14) (0.05)Food Shortage 0.092 0.030 0.173 0.061 -0.135 -0.047

(0.11) (0.03) (0.12) (0.04) (0.22) (0.08)Shock (idiosyn.) -0.003 -0.001 -0.098 -0.036 -0.552∗∗∗ -0.197∗∗∗

(0.10) (0.03) (0.10) (0.04) (0.19) (0.07)Shock (geogr.) -0.246 -0.083 -0.091 -0.033 0.220 0.073

(0.19) (0.07) (0.13) (0.05) (0.24) (0.08)Shock (institut.) -0.096 -0.031 0.105 0.038 0.395 0.126

(0.13) (0.04) (0.10) (0.04) (0.31) (0.09)Shock (other) 0.014 0.004 0.855∗∗ 0.253∗∗∗ -1.056∗∗ -0.363∗∗∗

(0.27) (0.09) (0.35) (0.07) (0.42) (0.12)Dist to Road -0.007 -0.002 -0.011 -0.004 -0.099∗∗ -0.028∗∗

(0.01) (0.00) (0.01) (0.00) (0.05) (0.01)Dist to Road2 -0.000 0.000 0.003∗

(0.00) (0.00) (0.00)Dist to Popcenter 0.011 0.003 0.001 0.000 0.018 0.005

(0.01) (0.00) (0.01) (0.00) (0.01) (0.00)Dist to Popcenter2 -0.000 0.000 -0.000

(0.00) (0.00) (0.00)Precipitation -0.000∗ -0.000∗ 0.000 0.000 -0.001 -0.000

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)cons -0.052 -1.294∗∗ 0.094

(0.50) (0.51) (1.12)N 1,590 1,590 1,463 1,463 479 479

36

Non-Farm Entrepreneurship in Rural Africa: Patterns and ...ftp.iza.org/dp8008.pdf · Non-Farm...

Documents

Transcript of Non-Farm Entrepreneurship in Rural Africa: Patterns and ...ftp.iza.org/dp8008.pdf · Non-Farm...